Karpathy - and others - consider the pre-training knowledge as much a liability as an asset. If we could just retain the emergent reasoning and language capabil

Karpathy - and others - consider the pre-training knowledge as much a liability as an asset. If we could just retain the emergent reasoning and language capability without the hazy recollections the models would likely be stronger.